Identification and Automatic Correction of Ammonia Probe Offsets in Wastewater Treatment Plants

In modern wastewater treatment plants, there is an increasing use of advanced intermittent aeration process controllers, which allow significant energy savings while maintaining high discharge quality standards. The measurement of ammonia in the tank is the ideal parameter for controlling intermittent aeration, especially in plants with restrictive discharge limits for this parameter. However, ammonia measurement probes are subject to frequent calibration losses, known as “offsets”, which require urgent calibration to ensure representative measurements and correct operation of the process controller. The real-time identification of these offset events and the automatic implementation of effective countermeasures represent a significant challenge due to the complexity of the process and the interactions between measurements, controllers, and biological processes. This study presents an innovative system that uses a machine learning algorithm to identify offset events and distinguish them from other anomalies, such as high input loads. The system then automatically verifies the extent of the probe offset using the process controller and corrects the measurement autonomously, thus ensuring that representative data can be used for aeration calculations. The algorithm was trained on a dataset from 17 plants and then refined through an active learning approach on two test plants, on which it was put into operation. In the first three months of operation, the system achieved a recall of 77% and a precision of 100%, correctly identifying ten offset events and automatically correcting measurements within 24 hours. These results demonstrate the effectiveness of the proposed system in improving plant management, offering a promising solution to the challenges of probe calibration, reducing the need for manual intervention, and optimizing aeration control.